Programmatic Advertising: forewarning and avoiding hype-cycle failure.

(Authors)

Abstract

The emergence of new technologies has often been examined through their transition along

the hype-cycle. While this has been a useful approach, recent research indicates that not all

new technologies follow the pattern of the hype-cycle as originally envisaged by Gartner.

Programmatic Advertising (PA) is a multi-billion dollar business that uses web-based

technologies to deliver highly personalised adverts to prospective consumers in real time.

Despite its rapid growth it has received precious little scholarly attention. This study is

therefore of interest to PA system developers and adopters since most have little

understanding of its operation and limitations, and are poorly equipped to make informed

decisions about its adoption and use.

Through the construction of a Concept Map of the system and the development of four future

states of Programmatic Advertising development, consumer concerns over the ethical usage

of data and the real return on investment are issues that are identified as requiring the

immediate attention of platform developers in order to mitigate the deleterious effects of

hype-cycle decline. The study proffers two alternative means by which the Programmatic

Advertising hype-cycle may develop, and unpacks the socioeconomic mechanisms by which

a loss of serendipity may occur in Programmatic Advertising systems.

Keywords: Programmatic Advertising, Hype-Cycle, Concept Map, Scenario Planning,

Sociotechnical Systems

Declarations of interest: None.

INTRODUCTION

Information Technologies (IT), in their many different guises, have had considerable impact

upon both business and society (Carlo, Gaski, Lytinnen and Rose, 2014; White, 2018).

However, while they afford new and more efficient ways of working and interacting, their

implementation is not straightforward and they may ultimately fail to meet expectations

(Adner, 2002; White, Gardiner, Prabhakar and Abdrazak, 2007; Schmidt and Druehl, 2008;

Chaffey and White, 2011; Mishra 2013; Thierer 2013; Sriram et al., 2015).

Programmatic Advertising (PA) is a relatively new implementation of IT that utilises large

data sets to disseminate deeply personalised marketing materials to target audiences

incorporating real-time pricing and bidding (Benady, 2015). Initially employed in web-based

advertising, the technique is finding application within film, television, apps, games and

loyalty schemes (Malthouse, Maslowska and Franks, 2018; Deng and Mela, 2018; Seitz and

Zorn, 2016; Gertz and McGlashan, 2016). The growth of PA has been rapid, with the market

estimated at being worth £960 million in the UK and almost $15 billion worldwide

(Lambrecht and Tucker, 2013; Aguirre et al., 2015; Benady, 2015; eMarketer, 2015). In 2015,

almost half of all digital adverts were traded programmatically and this is expected to rise to

over 80% in the near future (Benady, 2015). PA is said to provide organisations with a distinct

competitive advantage particularly when integrated with Customer Relationship Management

(CRM) systems (Lambrecht and Tucker, 2013; Benady, 2015; Hachen and Bardega, 2016;

Seitz and Zorn, 2016).

Despite the rising and rapid adoption of PA, there is very little scholarly literature that

examines this nascent phenomenon. The few extant studies are of value but tend to focus

upon discrete aspects of PA, including consumer responses, ethics, fraudulent web traffic and

PA’s application in television, but neglect to examine the system as a whole (Aguirre et al.,

2015; Busch, 2016; Fulgoni, 2016; Martinez-Martinez, Aguado and Boeyken, 2017;

Malthouse, Maslowska and Franks, 2018; Deng and Mela, 2018). This is troublesome since

PA is a complex sociotechnical system (Baxter and Sommerville, 2011; Trist, 1981) and

research is needed that studies its highly-interrelated elements (Benady, 2015; Seitz and Zorn,

2016; Brosche and Kumar, 2016; Gertz and McGlashan, 2016; Gangadharbatla et al., 2017).

Concerted effort is required, between practitioners and scholars, in order to define and

theorize PA (Gangadharbatla et al., 2017; Schwarz and Stensaker 2014). Since the approach

spans consumer behaviour, advertising, marketing, information technology, big data and

analytics, research into its opportunities and challenges should draw upon a similarly

ecelectic collection of disciplines and theoretical perspectives.

The ‘hype cycle’, developed by Gartner Inc., is an increasingly popular model that is used to

help researchers to analyse and forecast the evolution and commercial progress of

technologies (such as PA) in the marketplace (Dedehayir and Steinert, 2016; Jun, 2012a).

Hype cycle modelling has been adopted in numerous studies to develop a shared

understanding of a specific emerging technology and to determine consumer attraction and

diffusion patterns that can help to inform specific performative action (Van Lente, Spitters

and Peine, 2013). Seitz and Zorn (2016) argue that the significant hype generated by the PA

industry and the press has been responsible for its widescale unconscious adoption. This has

lead to the disruption of the traditional advertising industries eco-system which normally

comprises non-personalised mechanisms of designing and placing advertisements through

traditional media channels. In such cases target consumers are generically grouped according

to the rules of marketing segmentation (demographics, behavioural, psycographic) and

advertisments are designed around research that outlines how a particular product or service

may meet the needs and wants of the targeted group (Kotler and Armstrong, 2015).

The popularity of PA, evidenced by its rapid adoption and profusion among web searches,

indicates its importance as a subject of research. However, the recent emergence of criticisms

of PA’s capabilities suggest that PA is balanced upon the initial peak of the hype-cycle and

that rapid decline lies ahead. Seitz and Zorn (2016) concur and state that the PA industry is

now facing a period of inhibiting uncertainty and this prompts the aim of this research which

is to explore PA’s trajectory along the hype cycle.

This paper addresses the lack of research that examines the complexities of PA and proffers

the first step toward an understanding of the system as a whole. A comprehensive Concept

Map of the consituent elements of PA is constructed and used to instigate discussions with

expert programmatic practitioners about the tensions that exist within the system. By

uncovering the tensions that lie at the core of the PA system this paper moves beyond a

singular case study of practice and positions PA’s innovative disruption on the hype cycle. In

doing so, four future scenarios of PA development are generated comprising ‘Perfect

Algorithms’, ‘Ethical Limits’, ‘Negative Cost Advantage’ and ‘Fewest Platforms’. Inspired

by Dedehayir and Steinert’s (2016) analyses that challenge the accepted notion of the

inexorable ‘rise-fall-rise’ pattern of the hype-cycle, the study proffers two alternative means

by which the four future scenarios may manifest. This is a valuable critical interrogation that

informs future directions for system developers and technology adopters who may be guilty

of ‘blindly following the technological hype’ of this system that is increasingly generating

societal and economic concerns (Susi and Nicole, 2017). Its findings help PA platform

developers mitigate PA’s imminent descent into the hype cycle’s ‘trough of disillusionment’

by identifying specific mechanisms by which PA’s effectiveness may become eroded. While it

may not be possible to entirely eliminate the relapse that typically follows the hype of new

technology, attempting to ameliorate its effects so that the trough is neither as deep nor

perhaps as long lasting, would be a desirable outcome for PA developers, adopters and users

alike.

The paper is organized as follows: the next section reviews the concept of hype-cycles before

presenting the extant PA literature. The method of development of the PA Concept Map is

then detailed before its operation is discussed. The literature and the Concept Map are then

used to inform an examination of the tensions that are inherent in the PA system before four

future scenarios of its development are presented. These scenarios are then mapped onto the

‘Typical’ hype cycle and two variants, termed the ‘Concurrent’ hype cycle and the

‘Sequential’ hype cycle. The paper closes with concluding comments, statements of

limitations and suggestions for future research.

LITERATURE

HYPE-CYCLES

The concept of technology hype cycles was first proposed by Gartner (1995). Since then, they

have been studied in a range of contexts and across different technologies including fuel cells

(Konrad, Markard, Ruef and Truffer, 2012), hybrid cars (Jun, 2012a), voice over internet

protocol, gene therapy and superconductivity (Van Lente, Spitters and Peine, 2013), creative

arts (Abbasi, Vassilopolou and Stegioulas, 2017), additive manufacturing (Gartner, Maresch

and Fink, 2015), biomedical technologies (Boni, 2018), blockchain (Kewell and Ward, 2017)

and Corporate Prediction Markets (Womfram, 2015). Hype cycles generally conform to five

stages of expectation over time (Figure 1).

Stage 1: Innovation trigger: Awareness surrounding the novelty of new technology begins

to spread amongst users influencing early adopters to purchase/use the technology.

Organisations start to emerge with the hope of maximising the commercial advantage of

being the first to market (Van Lente Spitters and Peine, 2013). However during this phase,

while media attention could be high, some organisations experience a deficit in marketing

resources and subsequently risk failing to commercialise the technology at the right time (Jun

2016).

Stage 2: Peak of inflated expectation: Inflated by the hype generated from a variety of

media sources, this stage witnesses organisations investing and engaging in the technology

without clear strategic aims or objectives (Dedehayir and Steinert 2016). This stage in the

hype cycle is often associated with organisations and customers jumping on the ‘bandwagon’

(Dedehayir and Steinert, 2016) following the publication of ‘numerous initial success stories’

(Jun, 2016, 1414). Ultimately this leads to a peak in ‘optimism and exaggerated expectations’

regarding the technologies use and commercial viability (Van Lente Spitters and Peine, 2013,

1611).

Stage 3: Trough of disillusionment: This is a period of realisation and ‘realistic re-

adjustment’ where the media becomes more prone to generating negative news regarding the

failing application and/or commercial viability of the technology (Jun 2016, 1414). As Van

Lente, Spitters and Peine (2013, 1616) explain this stage in the hype cycle represents a

disappointment in the technology and is ‘marked by an abrupt collapse of positive

expectations’.

Stage 4: Slope of enlightenment: At this stage a more mature application and understanding

of the technology emerges, resulting in it becoming socially acceptable and performing to a

higher all-around standard (Gartner, 2018; Dedehayir and Steinert, 2016).

Stage 5: Plateau of productivity: This is the stage where commercial viability is proven and

broader applications and markets become available to the technology (Gartner, 2018; Jun,

2012a)

Plateau of Productivity

Time

Expectations

Slope of Enlightenment

Trough of Disillusionment

Peak Expectations

Innovation Trigger

Figure 1, The Hype-Cycle

The majority of literature makes reference to hype-cycles in the context of information

technologies, comprising discussions around ‘big data’ (Abbasi, Sarker and Chiang, 2016;

Bosch, 2016; Chen, Chiang and Storey, 2010), mobile communication (Adamauskas and

Krusinkas, 2017; Ozakazi and Barwise, 2011), e-business (Dainty, Leiringer, Fernie and

Hartty, 2017; Cihanek, Haseman and Ramamurthy, 2014; Au and Kauffman, 2005), e-

government (Bannister and Cnnolly, 2012), Web 2.0 (Bell and Loane, 2010), tabletop

computing (Bruun, Jensen, Kristenses and Kjeldskov, 2017), cloud computing (Willett, 2014;

Iyer, Krishnan, Sareen and Panda, 2013), online education (Mcpherson and Bacow, 2015),

social media (Roberts and Candi, 2014; O’Leary, 2011), healthcare (Reddy and Sharma,

2016), and the Internet of Things (Urquhart and Rodden, 2017).

Despite the conceptual usefulness of hype-cycles, much of the literature fails to adopt them as

a theoretical lens through which new technology adoption may be explored. For example,

some literature acknowledges and uses Gartner’s hype-cycle model (Boni, 2018; Kewell and

Ward, 2017; Bruun, Jensen, Kristenses and Kjeldskov, 2017; Urquhart and Rodden, 2017;

Stratopoulos, 2017; Reddy and Sharma, 2016; Bosch, 2016; Womfram, 2015; Willett, 2014;

O’Leary, 2011; Bell and Loane, 2010; Chen, Chiang and Storey, 2010; Wang, 2010;

Swanson and Ramiller, 2004; Ramiller and Swanson, 2003), whereas other studies make

only fleeting reference to hype-cycles (Adamauskas and Krusinkas, 2017; Dainty, Leiringer,

Fernie and Hartty, 2017; Gartner, Maresch and Fink, 2015;McPherson and Bacow, 2015;

Roberts and Candi, 2014; Xiatong, Kauffman, Yu and Zhang, 2014; Iyer, Krishnan, Sareen

and Panda, 2013; Bannister and Connolly, 2012; Ozakazi and Barwise, 2011; Au and

Koffman, 2005; Fichman, 2004), and both Ciganek, Haseman and Ramamurthy (2014) and

Nielsen and Fjuk (2010) merely refer to the general ‘hype’ that surrounds information

technology adoption.

Dedehayir and Steinert (2016) state that the hype cycle model has become popular model that

researchers have been used to critically evaluate technologies during the key stages of their

development. The model is praised for its ability to provide a useful framework to explain

and plot the adoption of technological innovations and critically evaluate the users'

expectations (Jun, 2012a). It is also suggested that it helps researchers take a more measured

view of disruptive technology by concentrating on the procedural aspects of the technology

whilst also considering the viewpoint of consumers and/or end users (Jun, 2012a). Hype

cycle studies are used in the literature in order to better understand diffusion patterns, but

they often follow a case study approach and are subsequently accused of being limited, only

generating findings that are relevant to a single product or service (Van Lente, Spitters and

Peine, 2013).

Much work has been done to theorise hype-cycles (Fenn and Reskino, 2009; Van Lente,

Spitters and Peine, 2013; Dedehayir and Steinert, 2016) and while they form useful

perspectives from which to view innovation processes and systems, they vary considerably

between contexts (Van Lente, Spitters and Peine, 2013). Dedehayir and Steinert (2016) offer

valuable critical insight into hype-cycles through their observation that while they may

conform to the pattern indicated in Figure 1, they more frequently manifest as a series of

peaks and troughs, and may even lack any form of recovery phase. They also add that the

pattern of technology expectation that underpins the hype-cycle may be different for the

different system actors.

PA Overview

A formal definition of PA is lacking due to a paucity of research and a great deal of

misunderstanding of its functioning (Whitmer, 2018; Alaimo, Kallinkos and Sess-Sforze,

2017). Fundamentally, it is a data-driven system that facilitates the real-time bidding for

advertising space to deliver personalized marketing materials to potential customers (Aguirre,

Mahr, Grewal, Ruyter and Wetzel, 2015; Benady, 2015; Funk and Nabout, 2016; Li, Yuan,

Zaho, Wang, 2017; Bush, 2016; Li. et al., 2017; Waesch, Rotberg and Renz, 2016; Gertz and

McGlashan, 2016; Kosorin, 2016). PA has radically altered the way that advertising is

undertaken (Li, Yuan, Zaho, Wang, 2017; Seitz and Zorn, 2016) and is capable of

considerably reducing the cost and risk of advertising (Lambrecht and Tucker, 2013; Benady,

2015; Bleier and Eisenbeiss, 2015; Aguirre et al., 2015). For example, The Economist used

PA to develop profiles of potentially suitable viewers by matching their reading preferences,

subscription data, web cookies and mobile app data. This enabled the publishers to

communicating real time messages that directly related to the specific individual interests

(finance, politics, social justice etc.) of every target customer (Globalwebindex, 2019). This

example outlines how PA enabled a fluid marketing communication campaign that was

individually targeted in terms of to whom, where and when it would appear and personalised.

Thus, via PA the Economist could take advantage of the use of ‘real time information’ and

‘opportunity creation’ to purchase and place the right advert in the right place at the right

time at an optimum price (Busch, 2016; Benady, 2015).

In brief, the PA system comprises several ‘platforms’ and actors. Data Management

Platforms (DMPs) profile customers from their browsing habits, purchase history and

personal preferences, typically from data stored as ‘cookies’. Other data may also be used

such as GPS location, current activities and weather conditions. For example, hotels that are

located near airports may use flight delay data to target stranded passengers with offers for

accommodation via their mobile phones (Gertz and McGlashan, 2016). Supply Side

Platforms (SSPs) manage the inventory of available advertising spaces – typically space on a

webpage but this varies depending upon the channel. Demand Side Platforms (DSPs) utilize

the DMP profiles to assess the ‘fit’ between the customer and the advertising materials of

participating organisations, then calculate the ‘value’ of that webspace and carry out the

auction-style bidding on behalf of participating organizations (Benady, 2015; Bush, 2016;

Kosorin, 2016; Schafer and Weiss, 2016.

The extant literature comprises predominantly practitioner articles that describe the benefits

of PA and predict its continued growth (Benady, 2015; Buch, 2016; Kosorin, 2016; Schafer

and Weiss, 2016; Seitz and Zorn, 2016). Comparatively little of this examines its

considerable complexity (Kosorin, 2016; Anderl, Schumannand and Kunz, 2016) and

questions over its actual effectiveness are beginning to emerge. For example, the literature is

punctuated with cautionary tales of costly mistakes (Benady, 2015; The Guardian, 2017; The

Telegraph, 2017), malpractice (Innovation in Magazine Media, 2016), risks (Seitz and Zorn,

2014), creative challenges (Weisbrich and Owens, 2016), confusion (Krefetz, 2016),

complexity (Benady, 2015; Anderl, Schumannand and Kunz, 2016), mistrust (Bleier and

Eisenbeiss, 2015) and contradiction (Benady, 2015; Aguirre, Mahr, Grewal, Ruyter and

Wetzels, 2015). In addition, technological advancements in web bots can produce fake page

impressions that distort metrics of customer views. It is estimated that 25% of video

impressions are in fact ‘viewed’ by bots and these fraudulent practices are costing US firms

around $4.5 million per hour (Fulgoni, 2016; Innovation in Media Magazine, 2016).

Adblocker technology is also developing rapidly and this adversely affects PA effectiveness

(Shiller, Waldfogel and Ryan, 2018; Turner, Shah and Jain, 2018). To counteract this,

organisations are employing methods of defeating the adblockers (Bashir, Arshad, Kirda,

Robertson and Wilson, 2018) and consequently an ‘arms race’ of blocker versus antiblocker

is escalating.

PA’s automated capabilities lead to its apparent cost effectiveness but also remove human

judgement from the process and this can result in improper advert placement (Benady, 2015;

Campaign Live, 2018). There have for instance been several cases where organisations have

withdrawn from PA platforms after their adverts had been displayed next to extremist

materials (The Guardian, 2017; The Telegraph, 2017). This highlights the need for marketers

to take greater care when utilizing PA platforms and not become seduced by the promises of

cost reductions (Schafer and Weiss, 2016). However, this may prove difficult because the

sheer technical complexity of PA is often beyond their ability to understand (Benady, 2015;

Seitz and Zorn, 2016; Gertz and McGlashan, 2016).

Organisations also need to be mindful of the loss of serendipity that may be encountered

through dynamically targeting customers which locks them into an ‘echo chamber’ of

exposure to repetitive adverts (Lambrecht and Tucker, 2013). Serendipitous experiences are

valuable elements of human learning and discovery but the argument of whether they can be

generated by information technologies remains moot (Andre, Teevan and Dumais, 2009;

Makri et al., 2014; De Gemmis et al., 2015; McCay-Peet and Toms, 2015; Erdelez and

Jahnke, 2018; Eirinaki, Gao, Varlarmis and Tserpes, 2018; Jain and Gupta, 2018; Kotkov,

Zhao, Konstan and Veijalainen, 2018). In order to provide more personalised adverts that are

relevant to the viewer’s current location, circumstances and needs, increasingly large and

personal data sets are required. However, this has the potential negative consequence of being

perceived as overly intrusive by prospective customers (Aguirre, Mahr, Grewal, Ruyter and

Wetzels, 2015; Van Doorn and Hoekstra, 2013). Data privacy is an increasingly sensitive

moral and legal issue (BBC, 2018), as evidenced by recent allegations of impropriety in the

US Presidential elections and Cambridge Analytics use of Facebook use data (Forbes, 2017;

The Guardian, 2018) and data privacy laws are constantly being revised to cope. Thus, if

organisations using PA continue to ignore the complexities of the system they may find

themselves wasting considerable amounts of funds, negating timely promotional

opportunities, isolating or scaring away existing and new consumers, devaluing their brand

equity or at worse flaunting legal requirments around data protection laws.

The considerable volume of practitioner literature discussed in this section that promotes PA,

coupled with a significant increase of the frequency of the term ‘Programmatic Advertising’

appearing in Google search results (see Figure 2), and its rapid and widespread adoption,

mirror the early phase of a technological hype-cycle (Dedehayir and Steinert, 2016; Van

Lente, Spitters and Peine, 2013). In addition, the recent appearance of articles that are critical

of PA’s actual efficacy (Funk and Nabout, 2016; Weisbrich and Owens 2016; Aguirre, Mahr,

Grewal, Ruyter and Wetzels, 2015; Bleier and Eisenbeiss, 2015; Seitz and Zorn, 2014), and

are summarized in Table 1, would suggest that the market has reached the point of ‘peak

expectation’ and may well be faced with the ‘trough of disillusionment’ (Figure 2). Indeed,

Seitz and Zorn (2016) concur and question whether PA’s rapid uncontested ‘hype cycle’ of

growth will result in the next .com crash (Seitz and Zorn, 2016).

1990-1991

1991-1992

1992-1993

1993-1994

1994-1995

1995-1996

1996-1997

1997-1998

1998-1999

1999-2000

2000-2001

2001-2002

2002-2003

2003-2004

2004-2005

2005-2006

2006-2007

2007-2008

2008-2009

2009-2010

2010-2011

2011-2012

2012-2013

2013-2014

2014-2015

2015-2016

2016-2017

2017-20180

10000

20000

30000

40000

50000

60000

70000

80000

90000

Google Search Results for "Programmatic Advertising"

Figure 2, “Programmatic Advertising” Articles

Table 1, Critical Literature

Year PA Challenge Article2014 Risks Seitz and Zorn (2014)2015 Expense Benady (2015)

Complexity Benady (2015)Mistrust Bleier and Eisenbeiss (2015)Contradiction Benady (2015)

Aguirre, Mahr, Grewal, Ruyter and Wetzels (2015)Improper Ad Placement Benady (2015)

2016 Malpractice Innovation in Media Magazine (2016)Creative Challenges Weisbrich and Owens (2016)Confusion Krefetz (2016)Complexity Anderl, Schummannand and Kunz (2016)

Seitz and Zorn (2016Gertz and McGlashan (2016)

Technological Fraud Fulgoni (2016)Innovation in Media Magazine (2016)

2017 Improper Ad Placement The Guardian (2017)The Telegraph (2018)

Technological Fraud Shiller, Waldfogel and Ryan (2018)Turner, Shah and Jain (2018)

2018 Improper Ad Placement Campaign Live (2018)

Managing this transition is particularly challenging for PA due to its inherent complexity, the

multiple stakeholders that are involved in its operation and the lack of contemporary research

and understanding. In order to enable PA platform developers to consider the wider

implications of technological decisions, and to enable existing and future adopters to make

informed decisions about the technology, this paper proffers a system map of PA in the form

of a Concept Map. This is used to identify the inherent tensions that exist within the system

that conspire to imbue it with a considerable degree of sociotechnical complexity – a feature

of other technologies that exhibit hype-cycles (Jun, 2012b). The tensions are used to compile

four future scenarios of PA that may contribute to its decline into the ‘trough of

dissilusionment’.

METHODOLOGY

Systems Dynamics (SD) is an approach to understanding systems that considers them in

terms of their elements and flows. Grounded in the field of Industrial Dynamics, Forrester

(1961) developed SD in order to model industrial management problems. SD maps are

popular tools for representing the dynamic nature of complex systems. Several types of

mapping techniques have been developed that suit specific applications and includes causal

loop diagrams, cognitive maps and concept maps, each of which could have been adopted for

this investigation (Schaffernicht, 2017; Georgiadis, Vlachos and Lakovou, 2005; Safayeni,

Derbentseva and Canas, 2005). This study uses Concept Mapping (CM) for its ability to

display important information that cannot be included in other techniques (Schaffernicht,

2017), represent knowledge of subject matter (Novak and Canas, 2008) and highlight the

dynamic relationships between events (Safayeni, Derbentseva and Canas, 2005) - see

Safayeni et al. (2005) for a detailed review of the origins and development of CM. CM has

been used in a variety of circumstances, most often education (see for example Horton,

McConney, Gallo, Woods, Senn and Hamelin, 1993) but also in the investigation of social

media (Moreno, Kota, Schoohs and Whitehill, 2013), consumers and marketing (Joiner,

1998), organizational culture (Kolb and Shepherd, 1997) and, apposite to this study, as a

research instrument in its own right (Kinchin, Streatfield and Hay, 2010; Joiner, 1998).

The review of the PA practitioner and academic literature was used to inform the

development of the concept map shown in Figure 2. In the corresponding description of the

concept map the following conventions are used: activities or the outcome of events are

indicated in the diagram by arrows and described in the text using the convention

‘description’, the system variables are indicated in the diagram by boxes and in the

discussion below by Capitalised Phrases. This literature-derived concept map provided the

basis upon which discussions of the operations of PA and of its inherent tensions were based

that culminated in the generation of four future scenarios of PA development. In total,

discusions of around 2 hours duration took place with five marketing scholars and five web-

based marketing professionals (Denscombe, 2010; Fox, 2009). Each academic participant

was a Senior Lecturer, Reader or Professor in their respective field and had held their post for

a minimum of five years. Each of the expert practitioners had direct experience of developing

and managing Programmatic Advertising systems or platforms. The identity of individuals

and their respective institutions is not disclosed (Duclos, 2017; Babbie, 2009).

Conversations with the participants were initiated with the request to “Explain the

Programmatic Advertising system”. The subsequent discussions were unstructured in order to

let the themes develop organbically (Fetterman, 2010), typically taking the form “What are

the challenges within the Programmatic Advertising system?” and this data was used to

inform the detailed development of the system map and the identification of its tensions. The

elements and characteristics of the PA system were captured using instantaneously-sampled

field notes (Paolisso and Hames, 2010) and included verbal descriptions and ‘napkin

sketches’ that were drawn by the participants to explain elements of the PA system. The

concept map was compiled using InsightMaker (https://insightmaker.com) and may be

viewed or freely copied for further development (available at

https://insightmaker.com/insight/60224/Programmatic-Marketing). The final concept map

and the discussions of its operation were member validated by two PA technical staff

(Sandelowski, 1993).

THE PA CONCEPT MAP

Beginning in the upper left quadrant of Figure 2 there is assumed to be a Consumer Demand

for a product or service that results in a ‘web search’ being conducted that influences the

number of Websites Visited. This in turn results in a number of different Adverts Seen and

contributes to the consumers Browsing History that is stored in the form of browser

‘cookies’. The Adverts Seen may result in a product or service being ‘wanted’ in which case

the consumer would proceed to Click & Buy. Adverts that are ‘seen’ are registered as Click

Through and those that are ‘ignored’ are counted as Not Wanted. Both Click Through and

Not Wanted results initiate a ‘repeat search’ or the end of web searching. External Reviews,

such as Tripadvisor, provides ‘data’ that influences the Consumer Perceptions, as does their

own ‘experiences’ of searching for and purchasing goods and services. These are

instrumental in determining the consumer’s Trust in Product and Trust in Provider that, in

turn, influence their browsing habits.

Click & Buy and Click Through generate ‘data’ that may be analysed and thereby potentially

contribute to better understanding of website and advert effectiveness, as well as consumer

preferences, and contribute to the Quality of Web Metric Analysis. They may also ‘stimulate

new demand’ in consumers. Adverts Seen that are subsequently ‘ignored’ may not provide

such data, depending upon the sophistication of the web systems employed. Click & Buy

would result in ‘demand fulfilled’ and may fully or partially reduce Consumer Demand. Click

& Buy is a ‘sale’ that increases the Advertiser Income, which may also increase that

advertiser’s ability to offer a ‘competing bid’ and thereby raise the Bid Price. The highest

‘competing bid’ would set the Bid Price and the ‘winning bidder’ would become a function

of the ‘Advertiser Filter’ whereby the ‘winner’s advert’ then becomes one of the Adverts

Seen by the consumer via their web browser.

The Bid Price also influences the ‘revenue’ and raises the Website Owner Income. This

enables the owner of the website(s) to make an ‘investment’ in Website Development,

informed by ‘website design’ suggestions based upon the Quality of Web Metric Analysis,

that improves the Website Effectiveness. This investment is realised through higher Web

Page Value and higher Bid Price. Successful website development and improvement

improves its ‘attractiveness & retention’ properties and thereby affects the Websites Visited

by the consumer and their resultant browsing behaviour. The Quality of Web Metric Analysis

also influences the ‘viewer-advert matching’ and thereby affects the Website Valuation

which, in turn, creates a demand that influences the Bid Price.

The platform software provides the dashboard through which web owners and advertisers

may access web metric data. ‘Revenue’ from successful Bid Price enables platforms to make

‘investment’ in Algorithm R&D that improves the Matching Accuracy and thereby improves

the Quality of Web Metric Analysis. Increased Platform Income also enables higher

‘investment’ in Marketing Expenditure, further ‘advertising’ and a greater Number of Broker

Dashboard Users. This in turn further increases the Platform Income through dashboard

‘rental charges.

Figure 2, Concept Map of the Programmatic Advertising System

ANALYSIS & DISCUSSION

Tensions

This section reviews the tensions that are inherent within the PA system, based upon the

issues that have been highlighted within the literature review coupled with the interviews

with expert scholars. The issues of ‘algorithm accuracy’, ‘data ethics’, ‘fraudulent traffic’,

‘non-human judgement’ and ‘loss of serendipity’ that are recognised within the extant

literature are addressed in turn and their resultant effects are discussed (shown in Table 2). In

addition to this, two further mechanisms by which a loss of serendipity can occur are

identified. These have not yet been recognised within the literature and their effects are also

examined.

Tension Origins

Algorithm Accuracy Bush (2016), Kosorin (2016), Schafer and Weiss

(2016), Benady (2015)

Data Ethics BBC (2018), The Guardian (2018), Forbes (2017),

Aguirre, Mahr, Grewal, Ruyter and Wetzels (2015),

Van Doorn and Hoekstra (2013)

Fraudulent Traffic Bashir, Arshad, Kirda, Robertson and Wilson (2018),

Shiller, Waldfogel and Ryan (2018), Turner, Shah

and Jain (2018), Fulgoni (2016), Innovation in Media

Magazine (2016)

Non-Human Judgement Campaign Live (2018), The Guardian (2017), The

Telegraph (2017)

Loss of Serendipity Lambrecht and Tucker (2013), Concept Map

Analysis

Table 2, Programmatic Advertising System Tensions

Tension 1 - Algorithm Accuracy

The key attribute of PA lies in its ability to match adverts with potential consumers that are

viewing a webpage or are exposed to a digital advert through film or television. It follows

that the more ‘accurately’ that this matching can be carried out then the more effective the

advert will be: that is, the viewer is more likely to engage in the service or will procure the

product that is being offered. The ability of the algorithm to achieve this end is dependent

upon two key factors, the accuracy of the algorithm itself and the completeness of the data

that is processed by the algorithm: the term ‘complete data’ is used to infer data correctness,

timeliness and relevance. Both the accuracy of the algorithm and the completeness of data are

necessary for ‘perfect’ matches to be identified.

The improvement of matching algorithms is fundamentally driven by the ability of a platform

developer to fund research and development or procure a knowledge base such as patents,

software or expert individuals. It can be seen that any platform that possesses a more

effective matching algorithm will have obtained a distinct competitive advantage. This

would, in turn, result in the acquisition of more organisations that utilize their services and a

resultant rise in competitive bid prices, and enable the increase of platform charges.

Collectively this results in greater income for the platform broker and thereby enables further

investment in algorithm research and development. Acquiring larger data sets in order to

improve the algorithm matching capability leads to the tension of Data Ethics.

Tension 2 - Data Ethics

The completeness of the data that is processed is dependent upon the availability and cost of

acquiring viewers’ browsing and contextual data. It therefore follows that any platform that

can acquire more complete data, and more data in general, would be in possession of a

distinct competitive advantage. This would, in turn, result in increased business, and therefore

increased income, that would enable the acquisition of more complete data. However, the

ability to acquire more complete data is moderated not only by cost but is also limited by

contemporary legislation and viewer perceptions of privacy invasion. This highlights a key

contradiction within the PA system: in order to improve the efficacy of viewer-advert

matching increasing amounts of data are required that, in turn, raise viewer concerns over

their digital privacy.

Tension 3 - Fraudulent Traffic

It is increasingly difficult to discern the real efficacy of PA due to the advent of purposeful

‘bots’ that create fake web traffic. In order to provide accurate performance metrics to

participating organisations PA platforms will need to develop mechanisms for preventing

bots from creating false traffic. Those platforms that are able to do this would then be in

possession of a distinct competitive advantage, further enabling the funding of bot-avoidance

mechanisms. It is likely that, in response, bot technologies would improve, thereby leading to

a continuous cycle of expenditure on development.

Tension 4 - Non-Human Judgement

Part of PA’s attraction lies in its automation. However, while this delivers perceived cost

benefits and is resource-friendly, the removal of human judgement can result in improper ad

placement. This is damaging, not only to the organisation whose advert has been placed, but

also to the PA platform that made the placement. The speed of automated PA trading means

that it is impractical for marketers to undertake a final ‘sense check’ of ad placement. Instead

it suggests that algorithm developers need to incorporate some form of digital environment

analysis in order to avoid ad misplacement. The technical feasibility of this is moot but the

costs of development would need to be shouldered. PA platforms that could develop this

capability would not necessarily possess a competitive advantage but would be able to

mitigate what is a significant competitive disadvantage.

Tension 5 - Loss of Serendipity

There is danger that PA can lead to viewers being repeatedly exposed to the same, or similar,

adverts. This lack of serendipitous experience leads to consumer weariness whereby adverts

do not just have little impact but they are completely ignored. In order to avoid this, broader

data sets need to be utilised, comprising personal and contextual data, that enable matching

algorithms to recognise and display appropriate offerings. PA platforms that can offer a

serendipitous advert experience would be in possession of a distinct competitive advantage.

There is however a need to balance the degree of serendipitous exposure with the exposure to

products and services that are known to be of current interest to the viewer in order to

maintain customer loyalty and income. The question of ‘how much serendipity is enough’ is

one that requires attention.

There are two other mechanisms that may decrease viewers’ serendipitous experiences that

are not mentioned within the literature review but are evident from the inspection of the

concept map (indicated in the lower left quadrant of Figure 1). First, ‘loss of serendipity b’

whereby a PA platform that gains a distinct competitive advantage would be in the position to

enable the adverts of its base of participating organisations to be displayed more widely than

those of competing PA platforms. Viewers would then be more likely to be exposed to the

range of adverts of organisations that utilise that PA platform. It is conceivable that this

would lead to increasing income for that platform and its partnering organisations so that they

could collectively invest in further algorithm development and data capture to further

increase their competitive advantage. Viewers would ultimately be ‘locked in’ to viewing the

adverts from the leading platform. Second, ‘loss of serendipity c’ whereby larger

organisations, with greater financial reserves, are able to outbid smaller organisations. This

would lead to viewers being presented with only those adverts that belonged to larger

organisations. Both of these situations would lead to a decrease in serendipitous experiences

for viewers.

PA Actor Perspectives

The individual tensions that beset PA may be considered to be of more immediate concern to

one of the three actor groups; comprising PA Adopters, PA Platform Developers and

Consumers. However, the sociotechnical complexity of PA means that each tension has some

cumulative effect upon the others. Figure 3 depicts the three actor groups and the tensions

that would appear to be of primary concern within each of their domains.

For example, PA Adopters are primarily concerned with the overall cost benefit of PA

compared to more traditional means of advertising. Part of the cost benefit analyses needs to

take account of the PA metrics, for example, in terms of the number of adverts that were

delivered to human (and not digital/fake) target consumers, and how many advert views

resulted in a purchase. This then becomes an issue for PA Platform Developers to address in

being able to provide reliable metrics. Consequently, an improvement in the cost benefit of

PA is likely to result in its wider adoption and thereby increased Consumers exposure to

programmatically generated adverts. This, in turn, drives a reduction in serendipitous

experiences that is countered by the increasing use of personal data. Consumer perceptions of

intrusiveness increase with the use of larger and more personalised data sets and this may

provoke avoidance of situations where PA is implemented. Conversely, curtailing the use of

larger and more personalised data sets in order to accommodate Consumer perceptions of

intrusiveness results in a reduction in the cost benefit of PA.

Overall, what this indicates is the complex and interrelatedness of the components of the PA

system. Furthermore, that the overall efficacy of PA is dependent upon the goals of each

actor being in harmony with the expactations and perceptions of the others. For instance, PA

Platform Developers need to take heed of Consumer perceptions of intrusiveness when

endeavouring to improve their ability to match prospective consumers with targeted adverts.

Similarly, PA Adopters must be mindful of the damaging effects that can be incurred through

improper, automatic advert placement, and not be lured by the promises of cost benefits

alone.

Figure 3, PA Actors and Tensions

FUTURE SCENARIOS

This section builds upon the discussions of the PA tensions to generate four future scenarios

of PA. Developed by Kahn (Kahn and Weiner, 1967) scenario planning encompasses a suite

of approaches for picturing possible futures (Ringland, 2010; Bradfield et al., 2005;

Meadows et al., 1992; Van der Heijden, 1996; Raskin et al., 1998; Huss and Honton, 1987).

The resultant scenarios contain information about a given situation or system so that they

may be used to guide decision-maker’s thinking and are particularly suited to fast-moving,

complex, technology-based phenomena (Harries, 2003; Alexander and Becker, 1978).

Scenario development may be based upon qualitative evidence or quantitative modelling, and

usually culminates in the generation of several scenarios (Schwartz, 1991; Wack, 1985). The

process may be highly structured or more organic in order to allow the expertise of evidence

to give rise to insightful imaginations of the future (Camponove, Debetaz and Pigneur, 2004).

Adopting an organic approach, using discussions with expert scholars, the four scenarios for

the PA system are identified as ‘perfect algorithms’, ‘ethical limits’, ‘negative cost

advantage’ and ‘fewest platforms’. These appear to be congruent with the view that PA is

experiencing a hype-cycle pattern of adoption. Understanding these future scenarios, and

PA Efficacy

Metrics

Human Judgement

Data Ethics

Serendipitous Experiences

Algorithm Accuracy

Cost-Benefit

PA Platforms

PA Adopters

Consumers

their intrinsic tensions, affords insight into PA that may enable developers and adopters to

ameliorate the negative consequences of entering the ‘trough of disillusionment’ phase of the

hype-cycle.

Scenario 1 ‘Perfect Algorithms’

The development of programmatic algorithms may continue to the stage where they offer

near perfect matching between viewers and adverts. That is, algorithms are capable of

exposing viewers to adverts for products and services that they habitually consume or are

likely to consume, and also to products and services that meet their current circumstances

even if the individual was not aware that they needed those products or services or that they

were even available. For instance, adverts for cheap, local accommodation are presented to

travellers that are about to be affected by impending flight delays.

Comprising the tensions of ‘algorithm accuracy’ and ‘data ethics’, this scenario requires

several assumptions to be met. First, the large, ‘live’ data sets that are required for such an

action are available. Second, those data sets are not prohibitively expensive to acquire and

process. Third, the acquisition of such large amounts of information is not prohibited by the

prevailing legislation in the countries where the activity is being practiced. Fourth, that the

acquisition of large amounts of highly personal information does not result in viewers

perceiving that it is some form of invasion of their privacy. Last, that all PA platforms

develop ‘perfect algorithms’ at or around the same time, else Scenario 4 would take effect.

The outcome of all PA platforms possessing ‘perfect algorithms’ would be that none of them

would benefit from this is a competitive advantage. All platforms would be able to provide

comparable matching of viewers with adverts and therefore there would be no advantage for

organisations to place their business with one platform instead of another. Without any

discernible performance advantage it is likely that PA platforms would then enter a price-

competitive market and this may ultimately lead to Scenario 3.

This scenario could also manifest in other technologies such as that used by the top two

companies in the Forbes 2018 list of the ‘World's Most Innovative Companies', which are

ServiceNow and Workday (Forbes, 2019). Both companies offer technological solutions to

the management of dispersed workforces and clients and are investing heavily in the

development of algorithms to predetermine client and workforce needs and thus build

predictive models that can decide incoming request. Fundamentally they are developing

programmatic systems that predict future workload request, workflow determination and

client needs, much like a PA systems does but without the bidding process for media

placement. Thus, the need for these organisations to develop algorithm accuracy in order to

appropriately predict future scenarios presents a similar dialectic challenges to avoid Hype

Cycle decline.

Scenario 2 ‘Ethical Limits’

Consumer sensitivity to the capture and utilization of large sets of highly personal data may

rise to the point where the access to further data becomes limited. This may occur through

changes to legislation, perhaps motivated by consumer lobby groups or other political

pressures. It may also occur in an alternative manner whereby organizations that use PA, and

by association are utilizing vast data sets, become shunned by consumers: a move that has

been replicated for example in the avoidance of organizations that are perceived to be at odds

with consumers’ values regarding slave labour and the environment.

Underpinned by the tension of ‘data ethics’, this scenario is dependant upon several

assumptions. Firstly, that the improved efficacy of PA remains dependent upon increasingly

large data sources. Also, that these increasingly large and complex data sets can be

practicably and cost effectively acquired. Finally, that legislation and consumer attitudes

toward data privacy remain constant.

The outcome of the emergence of an ‘ethical limit’ on the degree to which the acquisition and

processing of large data sets is tolerated would suggest that the motivation to develop

algorithms further would diminish. It is possible that some further refinement could take

place but it is envisaged that the majority of development would have taken place to take

advantage of new data types. The result of this is that PA platforms would be unlikely to be

able to develop any algorithm or data-driven competitive advantage. Consequently, SSP

providers may become the dominant players since they would control the webspace that was

available for ads to be placed upon. This may result in a situation where premium webspace

prices rise to the point where only large organizations with correspondingly large marketing

budgets may be able to afford to outbid competing organisations, leading to the tension ‘loss

of serendipity c’ and potentially Scenario 3.

This scenario also may manifest for any technology that relies upon big data. Returning to the

example of ServiceNow and Workday discussed in the previous section their ability to

predict future workload requests would be improved through the capture and use of

increasingly large and personalised data sets. Consequently, they may well be faced with

issues of transgressing perceptions of the ethical use of data.

Scenario 3 ‘Negative Cost Advantage’

It is very probable that programmatic algorithms will continue to improve in their

effectiveness at matching consumers with products and services. Assuming that such

facilities are made available to the majority of the market, at a cost that is not prohibitive,

then it is highly likely that organisations would enter a ‘bid war’ in order to take advantage of

the increased sales that may ensue. Thus, the cost of programmatic advertising to the

participating organisation would exhibit a rising trend.

It is also likely that fraudulent traffic would also increase, both in type and frequency. This

would require a concomitant investment in bot-avoidance systems and an increased cost that

would need to be factored into the programmatic platform charges. The presence of

fraudulent traffic along with improving, but less-than-perfect, programmatic algorithms, plus

the inefficiences caused by the non-human judgement of ad placement, all conspire to further

reduce the cost-effectiveness of PA.

Under these conditions it is foreseeable that the actual return on investment may drop to the

point where PA no longer offers a meaningful financial advantage over more traditional

forms of marketing; channels that are also better understood by marketers (Benady, 2015;

Seitz and Zorn, 2016; Gertz and McGlashan, 2016).

Scenario 4 ‘Fewest Platforms’

The first PA platform develop to be able to develop or closely approximate a ‘perfect

algorithm’ would have obtained a distinct competitive advantage. Assuming that this is not

prohibitively expensive to achieve, and that the large data sets that are necessary are also

affordable and are not abhorrent to consumers (tension of ‘data ethics’), then this is likely to

result in that platform becoming dominant within the PA marketplace.

If a single, or few, PA platform were to become dominant then this would propagate a

‘bidding war’ whereby organisations engage in aggressive bidding to ensure that their adverts

are presented to viewers that are ‘perfectly matched’ to become consumers. This scenario

may well result in consumers being faced with limited choices (tension of ‘loss of serendipity

b’) and participating organisations may find that the platform costs rise to the point where

Scenario 3 transpires.

DISCUSSION

While it is impossible to predict the future with any certainty, having developed several

possible future scenarios for PA, it is desirable to at least consider the potential for each to

occur. As Jun (2012a) noted, the phases of the hype-cycle may be offset for different system

actors. Our prognostications reflect this by suggesting that Consumers are most likely to have

concern over the ‘Ethical Limits’ scenario unfolding. In fact, this is something that has

already been recognised in practice (Aguirre, Mahr, Grewal, Ruyter and Wetzels, 2015; Van

Doorn and Hoekstra, 2013). Consequently, it would be logical to consider this scenario as

one that PA Platforms should address immediately.

Contrastingly, PA Adopters are more likely to be concerned with the Negative Cost

Advantage scenario. However, this is a situation that would be exacerbated by any adverse

effect caused by the ‘Ethical Limits’ scenario unfolding in tandem. Therefore, this also

suggests that while PA Platforms should work towards improving the reliability of

performance metrics, this must not be done at the expense of ignoring current Customer

issues of data privacy and feelings of intrusiveness.

The development of increasingly accurate algorithms is undoubtedly an activity that is of

interest to PA Platforms since it is one way in which they may be able to assert a competitive

advantage. However, since this is constrained by ‘Ethical Limits’ and also inhibited by the

difficulties of imbuing information systems with true serendipitous capabilities, it would not

appear to be an issue that is of immediate concern. Similarly, while the ‘Fewest Platforms’

scenario would appear to be of concern to all three PA system actors, through reducing

market diversity, this situation would appear to be a long-term effect of the efficacy of PA as

a whole. Consequently, both of these scenarios are comparatively long-term situations that

may or may not become realised. Based upon this putative rationale, Figure 4 presents our

interpretation of the likelihood of each of the four scenarios causing, or contributing, to the

decline in PA utilisation according to the accepted or ‘Typical’ interpretation of the hype-

cycle.

Perfect Algorithms

Fewest Platforms

Negative Cost Advantage

Ethical LimitsExpectations

Figure 4, Navigating the Traditional Hype-Cycle

Responding to Dedehayir and Steinert’s (2016) observation that not all technologies conform

to the ‘Typical’ hype-cycle as stated by Gartner, we proffer two alternative interpretations of

the future development of the PA hype-cycle. Figure 5 depicts a ‘Concurrent’ hype-cycle

whereby it is assumed that the challenges of the four future scenarios are addressed within the

same relative time frame. Successfully tackling the challenges of each of the four future

scenarios may well enable a more rapid and/or greater degree of recovery of the technology

into the ‘plateau of productivity’ phase.

Figure 6 depicts our interpretation of a ‘Sequential’ hype-cycle whereby the challenges of the

four future scenarios are addressed in turn. We conjecture that the successful amelioration of

a specific set of issues may initiate a new phase of increased interest and adoption of the

technology. This may be particularly true if the problems that are addressed are those that are

significant to a new actors within the PA system. For instance, addressing concerns over data

privacy may lessen consumer concerns and thereby spark a renewed interest in PA adoption.

Similarly, improvements in data reporting may attract new partners to existing platforms, or it

may stimulate the entrance of new PA platform providers to the market. Additionally,

improved algorithms that reduce the instances of adverts being placed next to inappropriate

materials may rekindle organisations’ confidence in PA technologies.

Expectations

Perfect Algorithms

Fewest Platforms

Negative Cost AdvantageEthical Limits

Figure 5, Navigating the ‘Concurrent’ Hype-Cycle

Figure 6, Navigating the ‘Sequential’’ Hype-Cycle

Whether PA conforms to a ‘typical’ hype-cycle, or one of the alternative patterns that have

been presented, remains to be seen. However, each would be benfitted by the whole or partial

resolution of the issues that have been detailed in each of the four future scenarios.

Consequently, a discussion of the ways in which the issues that surround the adoption, usage

and development of PA systems is provided next.

Firstly, consumer and societal concerns over data privacy are a current issue that have already

had material effect on several sociotechnical systems and their usage. PA’s dependency upon

complex and personal data, and its probable use of ever larger data sets, suggests to us that

the real and perceived ethical use of consumer data is likely to be of immediate and ongoing

concern. Platform developers and participating organisations should be mindful of the

paradoxical requirements of rich data that enable real-time, personalised consumer targeting

but increase the possibility of alienating privacy-sensitive individuals.

Concerns have already been raised over the actual cost-effectiveness of PA. Increased

bidding competition, rising development costs and the need for improved and transparent

performance metrics would contribute to a further escalation of costs. Ultimately, the tensions

that are inherent within the PA system suggest that the return on investment may become

Perfect Algorithms

Fewest Platforms

Negative Cost

AdvantageEthical Limits

Time

Expectations

eroded to the point where PA offers no discernible advantage over other methods of engaging

with the consumer audience.

It is not uncommon for new technologies to cause a proliferation of providers and for this to

be followed by a period of consolidation whereby the market becomes dominated by a few

key players. It is likely that a similar situation will occur in PA. The dominant PA platform

providers may well be those that are able to offer more transparent metrics dashboards, better

bot-deterrence, data acquisition techniques, or demonstrably improved algorithm efficacy.

Finally, it is considered least likely that a ‘perfect’ PA algorithm could be developed. While

algorithms will undoubtedly continue to improve, their ability to provide a ‘human-like’

ability to consider the context and content of ads and their placement is, as yet, a desirable

future goal.

CONCLUSION

Programmatic Advertising (PA) is a multi-billion dollar technological development, that

demands interdisciplinary academic attention in order to capture and understand its rich and

impactful complexity. To date, very little scholarly attention has been paid to this apparently

hype-driven phenomenon. This study has addressed this gap by constructing a Concept Map

of the PA system and develop four future scenarios of its potential hype-cyle decline.

The paper makes several important contributions. First, through review of the emergent

literature that is critical of PA’s capabilities and an examination of the proliferation of the

term “Programmatic Advertising” across the internet, it is evident that PA is balanced upon

the initial peak of the hype-cycle. Examination of the tensions that are inherent within this

complex sociotechnical system, through the construction of a Concept Map of its constituent

elements, four future scenarios are developed that indicate the means by which PA may slip

into the ‘trough of dissilusionment’. These are arranged in order of likelihood of occurrence

and thereby afford some indication of the issues that participating organisations and platforms

developers should be mindful of. In particular, the real and perceived ethical use and

treatment of personal data is an immediate issue that platform developers and participating

organisation need to consider carefully.

Second, while there is a large body of work that acknowledges or utilizes the term ‘hype-

cycle’ most of this does so in a superficial manner. Recent research suggests that few

technologies actually develop in the manner that is depicted by Gartner’s hype-cycle model.

In response, this study proffers two alternative means by which Programmatic Advertising

technology may manifest. The ‘Concurrent’ model assumes that the simultaneous resolution

of the challenges that best PA adoption will ameliorate the effect of the ‘trough of

disillusionment’ by reducing its maximum decline and overall duration. Contrastingly, the

‘Sequential’ model assumes that challenges are addressed in order of need and that the

resolution of each may invoke a repetitive cycle of gains and losses, which ultimately lead to

the technology reaching the ‘plateau of productivity’.

Third, the loss of serendipity is a known issue for many marketing systems and comprises

exposing consumers to an ‘echo chamber’ of repeated product and service offerings. This

study advances our understanding of serendipity in PA systems by identifying two further

ways in which it may be eroded or lost. Firstly, through a reduction in the number of PA

platforms caused by the dominance of one, or few, PA platform providers in the marketplace.

Fourth, through the dominance of large organisations, with correspondingly large budgets,

that can induce and win a ‘bidding war’. This is an issue for platform developers that may be

reliant upon the development of PA algorithms in order to deliver serendipitous moments.

While the continued improvement of PA algorithms is clearly of their concern, they should

be mindful that other, economic, mechanisms may also conspire to adversely affect their

ability to deliver new and inviting offerings to prospective consumers.

Lastly, PA is a system that has, so far, largely delivered upon its promises, but is beginning to

be questioned by many of its users. However, it has become a ‘black box’ solution that is

poorly understood by most and this has prohibited its critical investigation. This study is the

first that provides a complete overview of the PA system through the generation of a detailed

Concept Map. In doing so, it affords practitioners a device that describes the major functions

of a PA system and enables them to make an informed decision about its adoption or

continued usage.

The study has some limitations not least of which is the confidence with which future

predictions may be made about complex sociotechnical systems. Also, while this study has

considered the perspectives of the different actors that are involved with PA systems it must

be recognised that these are the homogenised views of highly disparate groups. Despite this,

the issues of data ethics and actual return on investment are pressing matters that should be

carefully considered by PA adopters and addressed by PA developers. Also, PA is a complex

and evolving business and as such there is a proliferation of ways in which PA is

implemented. Consequently, the Concept Map that is developed in this study should be

regarded as a generic overview of the key elements that should be encountered in PA

systems.

Future research should endeavour to further our understanding of PA systems through the

examination of situation-specific PA applications. In particular, valuable contributions could

be made through more interdisciplinary research so that the interplay of the technical and

socioeconomic systems could be better understood. Research should also examine the effect

of privacy-sensitivity upon user’s attitude toward organisations and systems that depend upon

the acquisition and analysis of large data sets. Research is also needed that examines the

actual cycles or trajectories that are followed by nascent technologies. In particular, PA

appears to have conformed to Gartner’s original observation that new technologies

experience rapid increases in expectations before more critical commentaries are observed.

Research should monitor the development of PA, and examine other technologies, in order to

discern whether they conform to some other predictive pattern.

REFERENCES

Abbasi, M., Vassilopolou, P., Stegioulas, L. (2017) Technology Roadmap for the Creative Industries. Creative Industries Journal, 10(1), 40-58.

Abbasi, A., Sarker, S., Chiang, R.H.L. (2016) Editorial. Big Data Research in Information Systems: toward an inclusive research agenda. Journal of the Association of Information Systems. 19(2), i-xxxii

Adamauskas, S. Krusinkas, R. (2017) Optimal Timing for a Technology Shift in Mature Markets of Mobile Communication. Strategic Change, 26(4), 343-361.

Adner, R. (2002) When are Technologies Disruptive? A Demand-Based View of the Emergence of Competition. Strategic Management Journal, 23, 667–688.

Alexander, C.S. & Becker, H.J. (1978) The use of vignettes in survey research. Public Opinion Quarterly 42(1): 93-104.

Assink, M. (2013) Inhibitors of disruptive innovation capability: a conceptual model. European Journal of Innovation Management, 9(2), 215-233.

Au, Y.A., Kauffman, R.J. (2005) Rational Expectations, Optimal Control and Information Technology Adoption. Information Systems and e-Business Management, 3, 47-70.

Babbie, E. R. (2009) The Practice of Social Research. Wadsworth Publishing Company, Cengage Learning, Kentucky, USA.

Bannister, F. Connolly, R. (2012) Forward to the Past: lessons for the future of e-government from the story so far. Information Policy, 17, 211-226.

Baxter, G., & Sommerville, I. (2011). Socio-technical systems: From design methods to systems engineering, Interacting with computers, 23(1), 4-17.

BBC (2018) Cambridge Analytica taken to court over data storage. http://www.bbc.co.uk/news/technology-43501184. Accessed 5th May 2018.

Bell, J., Loane, S. (2010) ‘New-Wave’ Global Firms: Web 2.0 and SME Internationalisation. Journal of Marketing Management, 26(3-4), 213-229.

Benady, D. (2015) programmatic: Packed with potential and peril, Catalyst, Charted Institute of Marketing:13-19.

Bleier, A. & Eisenbeiss, N. (2015) The Importance of Trust for Personalized Online Advertising. Journal of Retailing 91(3):390-409.

Boni, A.A. (2018) Innovation Challenges and Opportunities in Biopharma, MedTech, Digital Medicine, and their Emerging Convergence: user & patient centric applications in the ‘Pharma 3.0 Business Model Paradigm’. Journal of Commercial Biotechnology, 24(1), 39-47.

Bosch, V. (2016) Big Data in Market Research: why more data does not automatically mean better information. GfK Research, 8(2), 56-63.

Bradfield, R., Wright, G., Burt, G., Cairns, G. & Van der Heijden, K. (2005) The Origins and Evolution of Scenario Techniques in Long Range Business Planning. Futures, 37, 795-812.

Brosche, K. & Kumar, A. (2016) Realtime Data Accelerates Online Marketing. In: Bush O (Eds) Programmatic Advertising The Successful Transformation to Automated, Data-Driven Marketing in Real-Time, Switzerland, Springer, pp. 221-232

Bruun, A., Jensen, K.E., Kristenses, D.H., Kjeldskov, J. (2017) Escaping the Trough: towards real-world impact of table-top research. International Journal of Human-Computer Interaction, 33(2), 77-93.

Bush, O. (2016) The Programmatic Advertising Principle. In Bush, O (Eds) Programmatic Advertising The Successful Transformation to Automated, Data-Driven Marketing in Real-Time, Switzerland, Springer, pp.3-16.

Campaign Live (2018) Campaign. https://www.campaignlive.co.uk/article/keeping-brands-safe-programmatic-world/1439273. Accessed 3rd May 2018.

Camponove, G., Debetaz, S. & Pigneur, Y. (2004) A Comparative Analysis of Published Scenarios for M-Business. Proceedings of the Third International Conference on Mobile Business.

Carlo, J.L., Gaskin, J., Lyytinnen, K. & Rose, G.M. (2014) Early vs Late Adoption of Radical Information Technology Innovations Across Software Development Organizations: an extension of the disruptive information technology innovation model. Information Systems Journal, 24, 537-569.

Catinean, I. & Candea, D. (2013) Charcteristics of the Cloud Computing Model as a Disruptive Innovation. Review of International Comparative Management, 14(5), 783-803.

Chen, H., Chiang, R.H.L., Storey, V.C. (2010) Business Intelligence and Analytics: from big data to big impact. MIS Quarterly, 36(4), 1165-1188.

Christensen, C.M. (1997) The innovator's dilemma: when new technologies cause great firms to fail. Harvard Business School Press: Boston, Massachusetts, USA.

Christensen, C.M. (2014) Disruptive Innovation. In: The Encyclopedia of Human-Computer Interaction, 2nd Ed, Soegaard, Mads and Dam, Rikke Friis (eds.). Available online at https://www.interaction-design.org/encyclopedia/disruptive_innovation.html

Ciganek, A.P., Haseman, W., Ramamurthy, K. (2014) Time to Decision: the drivers of innovation adoption decisions. Enterprise Information Systems, 8(2), 279-308.

Dainty, A. Leiringer, R., Fernie, S., Hartty, C. (2017) BIM and the Small Construction Firm:

a critical perspective. Building Research & Information, 45(6), 696-709.

Danneels, E. (2004) Disruptive Technology Reconsidered: A Critique and Research Agenda. Journal of Product Innovation Management, 21(4), pp.246-258.

Dedehayir, O. & Steinert, M. (2016) The Hype Cycle Model: a review and future directions. Technological Forecasting & Social Change, 108, 28-41.

De Gemmis, M., Lops, P., Semeraro, G. & Musto, C. (2015) An Investigation on the Serendipity Problem in Recommender Systems. Information Processing and Management, 51, 695-717.

Denscombe, M. (2010) The good research guide: For small‐scale social research projects (4th edn). Berkshire, UK: McGraw‐Hill Education.

Duclos, D. (2017) When Ethnography Doe Not Rhyme with Anonymity: reflections on name disclosure, self-censorship and storytelling. Ethnography, In-press.

Eirinaki, M., Gao, J., Varlamis, I. & Tserpes, K. (2018) Recommender Systems for Large-Scale Social Networks: a review of challenges and solutions. Future Generation Computer Systems 78(1):413-418.

eMarketer (2015) http.//bit/1BMJWIa. accessed 10th May 2018

Erdelez, S. & Jahnke, I. (2018) Personalized Systems and Illusion of Serendipity: a sociotechnical lens. ACM, New Brunswick.

Fetterman, D.M. (2010) Ethnography SAGE Publications.

Fichman, R.G. (2004) Going Beyond the Dominant Paradigm for Information Technology Innovation Research: emerging concepts and methods. Journal of the Association for Information Systems, 5(8), 314-355.

Forbes (2019) “The Worlds Most Innovative Companies” available at https://www.forbes.com/innovative-companies/#18eb20c81d65 [accessed 05/02/19]

Fox, N. ,(2009) Using interview in a research project. Yorkshire, UK: The NIHR Research Design Service.

Fulgoni G..M (2016) Fraud in Digital Advertising: A multibillion Dollar black hole: How marketers can minimise losses caused by bogus web traffic. Journal of Advertising Research 56(2):122-125.

Funk, B. &, Nabout N.A. (2016) Cross-Channel Real-Time Response Analysis. In Bush, O (Eds), Programmatic Advertising The Successful Transformation to Automated, Data-Driven Marketing in Real-Time, Springer, pp.141-152

Gartner (2018) available at https://www.gartner.com/technology/research/methodologies/hype-cycle.jsp {accessed 17th July 2018]

Gartner, J., Maresch, D., Fink, M. (2015) The Potential of Additive Manufacturing for Technology Entrepreneurship: an integrative technology assessment. Additive Manufacturing and Technology Entrepreneurship, 24(4), 585-600.

Georgiadis, P., Vlachos, D. & Lakovou, E. (2005) A system dynamics modelling framework for the strategic supply chain management of food chains. Journal of Food Engineering, 70, 351-364.

Gilbert, C. (2003) The Disruption Opportunity. MIT Sloan Management Review, 27-32.

Gertz, O. & McGlashan,. D (2016) Consumer-Centric Programmatic Advertising. In Bush, O (Eds), Programmatic Advertising The Successful Transformation to Automated, Data-Driven Marketing in Real-Time, Springer, pp.55-74

Globalwebindex (2019) available at https://blog.globalwebindex.com/marketing/programmatic-ad-buying/ [accessed 10/02/2019}

Hachen, N. & Bardega, S. (2016) Evolution of Digital Campaign Design and Management. In Bush, O (Eds), Programmatic Advertising The Successful Transformation to Automated, Data-Driven Marketing in Real-Time, Springer, pp.211-220.

Harries, C. (2003) Correspondence to what? Coherence to what? What is good scenario-based decision making? Technological Forecasting and Social Change, vol. 70, no. 8, pp. 797-817.

Horton, P.B., McConney, A.A., Gallo, M., Woods, A.L., Senn, G.J. and Hamelin, D. (1993) An Investigation of the Effectiveness of Concept Mapping as an Instructional Tool. Science Education, 77(1), 95-111.

Huss, W.R. & Honton, E.J. (1987) Alternative Methods for Developing Business Scenarios. Technological Forecasting and Social Change, 31(3), 219-238.

Innovation in Magazine Media (2016) Advertising: Digital Advertising Fraud,122-138.

Iyer, E.K., Krishnan, A., Sareen, G., Panda, T. (2013) Sectorial Adoption Analysis of Cloud Computing by Examining the Dissatisfier Landscape. The Electronic Journal Information Systems Evaluation, 16(3), 211-219.

Jain, A. & Gupta, C. (2018) Fuzzy Logic in Recommender Systems. In: Kacprzyk C (Ed) Studies in Computational Intelligence Volume 749, Springer.

Joiner, C. (1998) Concept Mapping in Marketing: a research tool for uncovering consumers’ knowledge structure association. Association for Consumer Research, 25, 311-322.

Jun, S. (2012a) A Comparative Study of Hype Cycles Among Actors Within the Socio-Technical System: with a focus on the case study of hybrid cars. Technological Forecasting & Social Change, 79, 1413-1430.

Jun, S. (2012b) An empirical study of users' hype cycle based on search traffic: the case study on hybrid cars. Scientometrics 91 (1), 81–99.

Kahn, H. & Wiener, A.J. (1967) The year 2000: a framework for speculation on the next thirty-three years. Macmillan, New York.

Kewell, B., Ward, P.M. (2017) Blockchain Futures: with or without Bitcoin? Strategic Change, 26(5), 491-498.

Kinchin, I.M., Streatfield, D. and Hay, D.B. (2010) Using Concept Mapping to Enhance the Research Interview. International Journal of Qualitative Methods, 9(1), 52-68.

Kolb, D.G. and Shepherd, D.M. (1997) Concept Mapping Organizational Cultures. Journal of Management Inquiry, 6(4), 282-295.

Konrad, K., Markard, J., Ruef, A. & Truffer, B. (2012) Strategic Responses to Fuel Cell Hype and Disappointment. Technological Forecasting & Social Change, 79, 1084-1098.

Kosorin, D. (2016) Introduction to Programmatic Advertising. Kosorin: Czechoslovakia.

Kotler, P. & Armstrong, G. (2016) Principles of Marketing Global Edition, Pearsons, London.

Kotkov, D., Zhao, Q., Konstan, JA. & Veijalainen, J. (2018) Investigating Serendipity in Recommender Systems Based on Real User Feedback. SAC, France.

Krefetz, N. (2015) Programmatic Buying: what it means, how it’s changing advertising. http://www.onlinevideo.net/2016/06/programmatic-buying-means-changing-advertising/. Accessed 18th June 2018

Lambrecht, A. & Tucker, C, (2013) When Does Retargeting Work? Information Specificity in Online Advertising, Journal of Marketing Research 50:561-576.

Lyytinnen, K. & Rose, G.M. (2003) The Disruptive Nature of Information Technology Innovations: The Case of Internet Computing in Systems Development Organizations. MIS Quarterly, 27(4), 557-595.

Makri, S., Blandford, A., Woods, M., Sharples, S. & Maxwell, D. (2014) Making My Own Luck: serendipity strategies and how to support them in digital information environments. Journal of the Association for Information Science and Technology, 65(11), 2179-2194.

Malthouse, E.C., Maslowska, E. & Franks, J.U. (2018) Understanding Programmatic TV Advertising, International Journal of Advertising.

Markides, C. (2006) Disruptive Innovation: In Need of Better Theory. Journal of Product Innovation Management, 23(1), pp.19-25.

McCay-Peet, L. & Toms, E.G. (2015) Investigating Serendipity: how it unfolds and what may influence it. Journal of the Association for Information Science and Technology, 66(7), 1463-1476.

McPherson, M.S., Bacow, L.S. (2015) Online Higher Education: beyond the hype cycle. Journal of Economic Perspectives, 29(4), 135-154.

Meadows, D.H., Meadows, D.L. & J. Randers, J. (1992) Beyond the limits: confronting global collapse, envisioning a sustainable future. Chelsea Green, Post Mills, Vermont.

Mishra, S. (2013) Helping additive manufacturing. Metal Powder Report, 68(4), pp.38-39.

Morena, M.A., Kota, R., Schoohs, S., and Whitehill, J.M. (2013) The Facebook Influence Model: a concept mapping approach. Cyberpsychology, Behavior and Social Networking, 16(7),

Nielsen, P., Fjuk, A. (2010) The Reality Beyond the Hype: mobile internet is primarily an extension of PC-based internet. The Information Society, 26, 375-382.

Novak, J.D. and Canas, A.J. (2008) The Theory Underlying Concept Maps and How to Consutruct and Use Them. Technical Report IHMC, Florida Institute for Human and Machine Cognition.

Ozakazi, S., Barwise, P. (2011) Has The Time Finally Come for the Medium of the Future? Journal of Advertising Research, March, 59-71.

O’Leary, D.E. (2011) The Use of Social Media in the Supply Chain: survey and extensions. Intelligent Systems in Accounting, Finance and Management. 18, 121-144.

Paolisso, M. & Hames, R. (2010) Time Diary versus Instantaneous Sampling: A Comparison of Two Behavioural Research Methods. Field Methods, 1-21.

Quinn, R.E. & Cameron, K.S. (Eds.). (1988) Ballinger series on innovation and organizational change. Paradox and transformation: Toward a theory of change in organization and management. New York, NY, US: Ballinger Publishing Co/Harper & Row Publishers.

Ramiller, N.C., Swanson, E.B. (2003) Organizing Visions for Information Technology and the Information Systems Executive Response. Journal of Management Information Systems, 20(1), 13-50.

Raskin, P., Gallopin, G., Gutman, P., Hammond, A. & Swart, R. (1998) Bending the curve: toward global sustainability. Stockholm Environment Institute, Stockholm.

Reddy, P., Sharma, B. (2016) Digitalisation: the future of healthcare. Journal of Business Management, 11, 126-135.

Ringland, G. (2010) The Role of Scenarios in Strategic Foresight. Technological Forecasting and Social Change, 77(9), 1493-1498.

Roberts, D.L., Candi, M. (2014) Leveraging Social Network Sites in New Product Development: opportunity or hype? Journal of Product Innovation Management, 31(S1), 105-117.

Sandstrom, C., Berglund, H. & Magnusson, M. (2014) Symmetric Assumptions in the Theory of Disruptive Innovation: Theoretical and Managerial Implications. Creativity and Innovation Management, 23(4), 472-483.

Safayeni, F., Derbentseva, N. & Canas, A.J. (2005) A Theoretical Note on Concepts and the Need for Cyclic Concept Maps. Journal of Research in Science Teaching, 42(7), 741-766.

Sandelowski, M. (1993) Rigor or rigor mortis: the problem of rigor in qualitative research revisited. Advances in Nursing Science. 16(2), 1-81.

Schaffenicht, M. (2017) Available from: https://www.systemdynamics.org/conferences/2007/proceed/papers/SCHAF239.pdf

Schafer, A. & Weiss, O. (2016) Understanding Demand-Side-Platfroms. In Bush, O (Eds), Programmatic Advertising The Successful Transformation to Automated, Data-Driven Marketing in Real-Time, pp.75-86. Switzerland, Springer.

Schmidt, G.M. & Druehl, C.T. (2008) When Is a Disruptive Innovation Disruptive? Product Innovation Management, 25, 347-369.

Schumpeter, J. (1942) Capitalism, Socialism and Democracy. Harper and Row: New York.

Schwartz, P. (1991) The art of the long view: paths to strategic insight for yourself and your company. Doubleday, New York.

Schwarz, G. & Stensaker, I. (2014) Time to Take Off the Theoretical Straightjacket and (Re-) Introduce Phenomenon-Driven Research. The Journal of Applied Behavioral Science, 50(4): 478-501.

Seitz, J. & Zorn, S. (2016) Perspectives of programmatic Advertising. In In Bush, O (Eds), Programmatic Advertising The Successful Transformation to Automated, Data-Driven Marketing in Real-Time, pp.37-54. Switzerland, Springer.

Shiller, B., Waldfogel, J. & Ryan, J. (2018) The Effect of Ad Blocking on Website Traffic and Quality. The RAND Journal of Economics 49(1):43-63.

Sriram, S., Manchanda, P., Bravo, M.E., Chu, J., Ma, L., Song, M., Shriver, S. & Subramanian, U. (2015) Platforms: a multiplicity of research opportunities. Marketing Letters 26(2):141-152.

Stevens, A., Rau, A. & McIntyre, M. (2016) Integrated Campaign Planning in a Programmatic World. In Bush, O (Eds), Programmatic Advertising The Successful

Transformation to Automated, Data-Driven Marketing in Real-Time, pp193-210. Switzerland, Springer.

Stratopoulos, T.C. (2017) Exercising Due Diligence in Studies of Duration of Competitive Advantage Due to Emerging Technology Adoption, Journal of Emerging Technologies in Accounting, 14(1), 27-35.

Susi, G. & Nicole, G. (2017) Does hype create irreversibilities? Affective circulation and market investment in digital health. Marketing Theory, 17(4),435-454.

Swanson, E.B., Ramiller, N.C. (2004) Innovating Mindfully with Information Technology. MISQuarterly, 28(4), 553-583.

The Guardian (2017) Guardian pulls ads from Google after they were placed next to extremist material.] Available from: https://www.theguardian.com/media/2017/mar/16/guardian-pulls-ads-google-placed-extremist-material. Accessed 2nd May 2015.

The Guardian (2018) The Cambridge Analytica saga is a scandal of Facebook’s own making. Available from: https://www.theguardian.com/commentisfree/2018/mar/21/cambridge-analytica-facebook-data-users-profit. Accessed 9th May 2018.

The Telegraph (2017) Google apologises over extremist video fears as Marks & Spencer pulls advertising, Available from: https://www.telegraph.co.uk/business/2017/03/20/marks-spencer-pulls-adverts-google-extremist-video-fears/. Accessed 2nd May 2018.

Thierer, A. (2013) Permissionless Innovation. Mercatus Center at George Mason University: USA.

Turner, J., Shah, R. & Jain, V. (2018) How Brands are Using Nonlinear Marketing to Address Consumer Resistance to Traditional Marketing Strategies. Journal of Brand Strategy 6(4):397-408.

Urquhart, L., Rodden, T. (2017) New Directions in Information Technology Law: learning from human-computer interaction. International Review of Law, Computers & Technology, 31(2), 150-169.

Van der Heijden, K. (1996) Scenarios: the art of strategic conversation. Wiley, New York.

Van Doorn, J. & Hoekstra, J.C. (2013) Customization of Online Advertising: the role of intrusiveness. Marketing Letters 24(4):339-351.

Van Lente, H., Spitters, C. & Peine, A. (2013) Comparing Technological Hype Cycles: towards a theory. Technological Forecasting & Social Change, 80, 1615-1628.

Wack, P. (1985) Scenarios: shooting the rapids. Harvard Business Re- view 63:139–150.

Wang, P. (2010) Chasing the Hottest IT: effects of information technology fashion on organization. MIS Quarterly, 34(1), 63-85.

Weisbrich, S. & Owens, C. (201) The Creative Challenge. In Bush, O (Eds), Programmatic Advertising The Successful Transformation to Automated, Data-Driven Marketing in Real-Time, pp.123-131. Switzerland, Springer.

White, G.R.T. (2018) Future Applications of Blockchain in Business and Management: a Delphi study. Strategic Change, Special Issue on The Future of Money and Further Applications of the Blockchain, in-press.

White, G.R.T., Gardiner, G., Prabhakar, G. and Abdrazak, A. (2007) A Comparison of Barcoding and RFIDTechnologies in Practice. Journal of Information, Information Technology and Organizations, 2, p119-132.

Whitmer, C. (2018) 5 Programmatic Mistakes that you’re probably making. Available from: (https://www.ama.org/publications/eNewsletters/B2BMarketing/Pages/5-Programmatic-Mistakes-You’re-Probably-Making.aspx). Accessed 5th June 2018.

Willett, M. (2014) Cloud-Based Email Adoption at Higher Education Institutions in South Africa. Journal of International Technology and Information Management, 23(2), 17-29.

Womfram, T. (2015) Have Corporate Prediction Markets Had Their Day? Foresight, Spring, 29-36.

Xiaotong, L., Kauffman, R.J., Yu, F., Zhang, Y. (2014) Externalities, Incentives and Strategic Complementarities: understanding herd behavior in IT adoption. Information System E-Business Management, 12, 443-464.

Yu, D. & Hang, C.C. (2010) A Reflective Review of Disruptive Innovation Theory. International Journal of Management Reviews, 12, 435-452.

Zhou, W. & Piramuthu, S. (2013) Technology Regulation Policy for Business Ethics: An Example of RFID in Supply Chain Management. Journal of Business Ethics, 116, 327-340.